Conversion rate: what it is, how to calculate it and how to increase it

What is conversion rate?

A conversion is an action that's valuable for your product. A purchase, a sign-up, a quote request or starting a trial are examples. You decide what counts as a conversion.

Conversion rate is the share of all visitors who take that action. Jakob Nielsen defines it as the percentage of users who take a desired action. Nielsen also stresses that a conversion doesn't have to be a sale.

In Google Analytics 4 you mark an action as a "key event" to count it as a conversion. According to Google, a key event is an event that measures an action that's particularly important to your business's success. Which event you mark decides what the rate measures.

So the sentence "our conversion rate is 3%" says nothing on its own. You also have to say which action and which audience you counted.

How to calculate conversion rate

The formula: conversion rate = conversions ÷ visits × 100. In Wikipedia's example a site gets 200 visitors a day. Of these, 30 sign up for the newsletter. That day's conversion rate is 15%.

Nielsen's example is an e-commerce site. In April it has 100,000 visitors and 2,000 purchases. The rate is 2%.

The denominator: sessions or users?

The hard part of the formula is the denominator. Google Analytics gives two different rates. The session rate divides sessions with the target action by all sessions. The user rate divides users who took the action by all users.

If the same person visits three times and buys once, the two rates split. By session it's 33.3%, by user 100%. According to Nielsen you can count unique visitors or repeat visits. Just count the same way every period.

In reports I always state the denominator, for example "session-based purchase rate, October, all channels". When someone asks about the figure two months later, nobody argues about what was measured.

One rate isn't enough: draw the funnel

The overall rate shows how big the problem is. To find where it is, you split the flow into steps: visit, product page, cart, checkout, purchase. This sequence, showing how many people are left at each step, is called a funnel.

Enter your own steps and numbers in the calculator below. It works out the overall rate and the loss at each step, and marks the step where the most people leave.

Conversion rate and funnel calculator

1. Step 1 name
   Step 1 count
2. Step 2 name
   Step 2 count
3. Step 3 name
   Step 3 count
4. Step 4 name
   Step 4 count

Overall conversion rate 2.00% (200 / 10,000 × 100)

Biggest loss: Product page → Cart. The loss rate at this step is 80.0%.

1. 1. Visit 10,000

   First step. The other rates are calculated from it.

2. 2. Product page 4,000

   Passed from the previous step: 40.0%. Lost: 6,000 people (60.0%). Versus the first step: 40.0%.

3. 3. Cart 800 Biggest loss

   Passed from the previous step: 20.0%. Lost: 3,200 people (80.0%). Versus the first step: 8.0%.

4. 4. Purchase 200

   Passed from the previous step: 25.0%. Lost: 600 people (75.0%). Versus the first step: 2.0%.

JavaScript is off, so the result is shown only for this sample funnel.

The tool collects and sends no data. The calculation runs in your browser.

For each step, enter how many people reached it. The first step is usually visits. The three sample funnels use numbers I made up and checked by hand; they aren't data from a real product.

You can check the e-commerce example by hand: 200 ÷ 10,000 × 100 = 2%. 4,000 people saw the product page and 800 of them went on to the cart. That's 20% passing, 80% lost.

From cart to purchase the pass rate is 200 ÷ 800 = 25%. So the real loss is between the product page and the cart. You can't see that from the overall rate.

Macro and micro conversions

Tim Neusesser of Nielsen Norman Group splits conversions in two. A macro conversion is an action that contributes directly to the business's main goal. A purchase, a submitted quote form or a new account belong here.

Micro conversions are linked to the main goal indirectly but happen more often. Adding to cart, starting checkout and a newsletter sign-up are examples. Neusesser divides them into two types.

• Process milestones: steps on the most likely path to a macro conversion. Starting checkout is the step before purchase.
• Secondary actions: actions that don't lead straight to a sale but can signal a future conversion, such as a newsletter subscription or downloading the app.

Macro rates change slowly. So Neusesser recommends measuring small design improvements with micro conversions. Each intermediate step in the calculator is a micro conversion of the process-milestone kind.

The article has an important warning for government and internal products. For mandatory tasks where users have no other option, conversion rate isn't a good measure. For things like filing taxes or requesting leave, look at error counts and support requests instead.

Tracking traps: the rate may be wrong

Before you touch a low rate with design, make sure the figure is right. In audits I've seen several times that the problem wasn't the interface but the tracking.

Bots and internal traffic

Google Analytics automatically excludes traffic from known bots and spiders. It does so based on Google's own research and the IAB's bot list. This setting can't be turned off, and you can't see how much traffic was excluded.

Bots that aren't on the list do count as visits and pull the rate down. If visits shoot up overnight but sales stay flat, that's the first thing I suspect.

Your team's own visits also inflate the denominator. In GA4 you can define internal traffic by IP address and exclude it with a filter. Once the filter is active, the excluded data is gone for good and can't be brought back. So try the filter in testing mode first.

Channel mix

The overall rate is a weighted average of the channels. If the channels' share of traffic changes, the overall rate can fall even when every channel improves. In statistics this is called Simpson's paradox: a trend seen in groups can disappear or reverse when the groups are combined.

The numbers below are an example I made up. In February both channels' rates go up. But because most traffic shifts to low-converting social media, the overall rate falls.

Channel	January	February
Search	5,000 visits, 4%, 200 sales	3,000 visits, 4.5%, 135 sales
Social media	5,000 visits, 1%, 50 sales	9,000 visits, 1.2%, 108 sales
Total	250 ÷ 10,000 = 2.5%	243 ÷ 12,000 ≈ 2.03%

So always read the rate split by channel.

Nielsen describes the same effect. An article about cats brought his site 4 times its normal traffic in one day. Sales stayed the same as on an average day, and the rate fell to almost a quarter of normal, because the visitors were outside the target audience.

Seasons and the measurement period

Nielsen recommends accounting for seasonal variations within the measurement period. Consumer sites usually sell more during the December holidays. Big sale weeks such as Black Friday have the same effect.

So compare November with last year's November, not with October. And don't measure the effect of a design change during a campaign week.

Conversions counted twice or not at all

If you count a conversion when the thank-you page loads, someone who refreshes the page can be counted twice. Someone who doesn't come back from the payment provider's page may not be counted at all. Once a month, compare the purchase count in analytics with the count in your order system.

What's a good conversion rate?

There's no answer that fits everyone, and I won't make up an average. The clearest source I found is Nielsen's 2013 article. According to Nielsen, a good conversion rate is usually in the 1% to 10% range.

The top of that range is ten times the bottom, and the figure is from 2013. Nielsen also writes that the rate varies with brand, price, the complexity of the sale and the commitment asked of the user. A product bought on impulse and a B2B service that needs management sign-off aren't measured on the same scale.

My advice: compare your own rate with your own history, for the same channel and the same period. If you make another site's average your target, you'll focus on the wrong step.

Low conversion rate: diagnose by symptom

A low rate has no single cause. I start the diagnosis from the symptom, because the symptom tells me where to look. These are the cases I meet most often in audits.

The rate dropped suddenly after a release

Likely cause: A tracking tag may have been removed, or a bug may have appeared in the flow. For example, the checkout step doesn't work in one browser.

Diagnosis: Match the day of the drop with the release date. Compare the purchase count in analytics with the order system. Then go through the flow yourself, end to end, on different devices.

Visits went up, sales stayed flat

Likely cause: The visitors may be outside the target audience, or analytics doesn't recognise some bots. In that case the page isn't broken; only the denominator has grown.

Diagnosis: Split the rate by channel and landing page. Lots of very short sessions from a single source make bots more likely.

Low on mobile, normal on desktop

Likely cause: Touch targets may be small, or a form field opens the wrong keyboard. Sometimes a bug shows up only on mobile.

Diagnosis: Split the funnel by device and see which step has the biggest gap. Try that step on a real phone. Compare it with the touch target rules.

Many add to cart, few buy

Likely cause: The shipping cost appears at the last step, or an account is mandatory. In Baymard's list, updated in September 2025, the most common reason after "just browsing" is high shipping, taxes and extra costs (40%). A forced account comes fourth at 18%.

Diagnosis: Add every checkout step to the funnel and see where the loss grows. The checkout page pattern and cart abandonment reasons pages are the checklist for this step.

Many open the form, few submit it

Likely cause: The form asks for fields it doesn't need, or the error messages don't say what to do.

Diagnosis: Look at abandonment data by field. Fill the form in wrong on purpose and read what the error says. Compare it with the form design and error message patterns.

The product page gets many views, few add to cart

Likely cause: Price or delivery information may be missing. The main button isn't visible, or the product isn't what people expected.

Diagnosis: Numbers don't explain this step; you have to hear the reason from users. Run a small usability test with a few people and ask why they left the page.

Every step looks reasonable, but the overall rate stays low

Likely cause: The problem may be the offer. Design doesn't fix a price or a product that doesn't fit the audience.

Diagnosis: Collect the objections the sales team and customer service hear most often. As a designer I don't enjoy saying this, but it's the honest answer.

The most common UX causes

If the tracking is right and the offer is sound, the loss usually comes from one of these problems:

• The total appearing only at the last step. Users leave the moment they see the price.
• Requiring an account before purchase. If continuing as a guest isn't visible, losses rise.
• Forms that ask more than they need. According to Nielsen, every question removed from a form raises the completion rate.
• Steps that don't work, or work badly, on mobile.
• Empty screens and error messages that don't tell users the next step.
• Details that shake trust. Vague return terms or an unfamiliar payment page make users stop.

There's also the opposite direction. A hidden subscription, a pre-ticked box or a hard-to-cancel flow can raise the rate in the short term. These count as dark patterns. The rate you gain comes back later as refunds and complaints.

How to increase conversion rate

I never change the order. Each step builds on the result of the one before.

1. Verify the tracking. Filter out bots and internal traffic, and compare analytics with the order system.
2. Draw the funnel. Split the main flow into steps, and each step by channel and device.
3. Focus on the biggest loss. The step the calculator above marks is usually the starting point.
4. Find the cause. The number shows where, not why. A heuristic evaluation and user testing fill that gap.
5. Make one change at a time. If you change ten things at once, you can't tell which one worked.
6. Confirm with an A/B test. Measure whether the change really made a difference.

Nielsen adds one more warning: optimise conversion, don't maximise it. The cost of raising the rate by one more point can be bigger than what it brings in.

Confirming with an A/B test

In an A/B test, visitors are split at random into two groups. One group sees the old version, the other the new one, and the two groups' rates are compared. The real question is whether the difference is down to chance.

You can check that with your own numbers in the A/B test calculator. The same page also works out how many visitors each group needs before you start the test. Fix the sample size up front, and don't stop the test early because you peeked at the result.

If your traffic is low, an A/B test can take weeks. Then I make the change and compare the period before and after, for the same channel and season. That's weaker evidence, and I say so plainly in the report.

My note

The case I see most in audits is a team that tracks the overall rate every week but has never drawn the funnel. When the rate drops, the homepage gets redesigned. Yet the loss is usually at another step, sometimes in a single field of the checkout form.

The second common case is two teams discussing the same rate with different denominators. Marketing counts sessions, the product team counts users. Half the meeting goes on arguing about what the figure is.

When I look at a conversion problem, I check the tracking first, then the funnel, and the screens last. UX Audit Pro covers up to 10 flows across the whole app and gives you a prioritised list of fixes. If you'd like to understand the process first, the UX audit guide walks through the steps one by one.